results lost?

2013-03-06, 13:06

My machine was testing a 100 million digit nuber for nearly one year and did some LL-D and smaller LL tests as well (mprime 26.6 Ubuntu 12.04 32bit intel i5 650 @ 3.20). I had do do a new install. So I copied the program folder to Flash drive. I ran the program from there (mprime27.9 64bit Scientific linux live system). As well as on a Ubuntu 12.10 64bit live system. Id did bechmark tests 32 vs. 64 bit. etc. copied more than once from hdd to flash and vice versa. Always kept the newer files... Suddenly my expected completion date was 10 years farther away. I think it was because of a file permissions issue. sudo does not have the power it used to have any more in 12.10. http://www.mersenneforum.org/images/icons/icon8.gif I had to change permissions by chmod command. So I think mprime started from scratch. I think I overwrote the old files. What about files mprime does not have write permission for? Does it create a new one with the same name? That's all I've got:

/mprime/p332199319
/mprime/p01C0989
/mprime/local.txt
/mprime/p332199319.bu
/mprime/results.txt
/mprime/p01C0989.bu
/mprime/prime.log
/mprime/p01C0989.bu2
/mprime/p332199319.bu2
/mprime/prime.txt
/mprime/worktodo.txt
/mprime/results64bit.txt
/mprime/p3W73763
/mprime/p3W73763.bu
/mprime/p3W73763.bu2
/mprime/mprime
/mprime/whatsnew.txt
/mprime/undoc.txt
/mprime/stress.txt
/mprime/readme.txt
/mprime/license.txt

Recovery possible? Thank You for help, Obscure

Brian-E · 2013-03-06, 17:33

Your data save files from that list, which you need to continue the tests, are p332199319 (I think this one is the 100M digit LL test), p01C0989, p3W73763. The same names with .bu and .bu2 extensions are normally slightly older backup versions of the data save files from those tests.

Of course, everything now depends on when those data files were made, and whether they've been overwritten by other versions when your tests restarted. So you need to look at the dates and times of last alteration: use the -l extension of the ls command ("ls -l") in Linux to show these times.

In worktodo.txt you can see what is currently being worked on (or was being worked on the last time mprime was running, assuming it had access to the file then). You need to make sure that this file contains what you want it to before running mprime.

Tip: make copies of what you have and store them elsewhere before you do anything else, certainly before you start mprime again!

Xyzzy · 2013-03-06, 18:00

Quote:

Tip: make copies of what you have and store them elsewhere before you do anything else, certainly before you start mprime again!

obscure · 2013-03-07, 13:54

Sorry, no older backups found. I should have stopped immediately after seeing something went wrong since the bu2 file is overwritten one hour later.

I hoped, some results were stored on the server, not to loose all of the work...
What if I unreserve an Exponent, is my work lost then?
I was not aware that root cannot read or execute a file but overwrite it as I think this is what really happened.

Thank You for help,

Obscure

Brian-E · 2013-03-08, 11:10

This must be so frustrating for you.

No, the server simply keeps track of how far through the work you are as reported by your machine, and as for the data processed in your test it only stores the final result. Even then, in the case of LL tests, the server in fact only stores the last 64 bits of the final residue. The data files on your own machine contain the full residue throughout the duration of the test, and this residue is megabytes long.

If you unreserve an exponent on the server without "telling" mprime on your machine (worktodo.txt is the master file here), mprime will continue the work. This is not, of course, a good idea: the exponent will at some point be assigned to someone else and you will be doing work which is not yours to do. If you unreserve it more cleanly using the menu system in mprime, or altering the worktodo.txt yourself after manually unreserving on the server, then I think mprime will leave the data files alone (but back them up first to be sure!): but again, if you've unreserved the exponent on the server you should not be continuing the work anyway.

Xyzzy · 2013-03-08, 11:25

You are not the first to lose work, so please do not feel too bad!

We keep a save file for every 1e6 iterations.

http://www.mersenneforum.org/showthread.php?p=120849

2013-03-06, 13:06	#1
Unregistered 2×13×263 Posts	results lost? My machine was testing a 100 million digit nuber for nearly one year and did some LL-D and smaller LL tests as well (mprime 26.6 Ubuntu 12.04 32bit intel i5 650 @ 3.20). I had do do a new install. So I copied the program folder to Flash drive. I ran the program from there (mprime27.9 64bit Scientific linux live system). As well as on a Ubuntu 12.10 64bit live system. Id did bechmark tests 32 vs. 64 bit. etc. copied more than once from hdd to flash and vice versa. Always kept the newer files... Suddenly my expected completion date was 10 years farther away. I think it was because of a file permissions issue. sudo does not have the power it used to have any more in 12.10. http://www.mersenneforum.org/images/icons/icon8.gif I had to change permissions by chmod command. So I think mprime started from scratch. I think I overwrote the old files. What about files mprime does not have write permission for? Does it create a new one with the same name? That's all I've got: /mprime/p332199319 /mprime/p01C0989 /mprime/local.txt /mprime/p332199319.bu /mprime/results.txt /mprime/p01C0989.bu /mprime/prime.log /mprime/p01C0989.bu2 /mprime/p332199319.bu2 /mprime/prime.txt /mprime/worktodo.txt /mprime/results64bit.txt /mprime/p3W73763 /mprime/p3W73763.bu /mprime/p3W73763.bu2 /mprime/mprime /mprime/whatsnew.txt /mprime/undoc.txt /mprime/stress.txt /mprime/readme.txt /mprime/license.txt Recovery possible? Thank You for help, Obscure

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Lost assignment?	JuanTutors	PrimeNet	31	2010-06-30 20:04
Lost numbers ?	Nikolaus	Homework Help	1	2010-03-14 13:03
Lost work	M0CZY	Information & Answers	0	2007-07-26 14:05
Lost avatars.	Xyzzy	Forum Feedback	26	2007-03-23 01:53
Uh, Lost	Darkfire001	Software	2	2002-10-11 02:08

2013-03-06, 17:33	#2
Brian-E "Brian" Jul 2007 The Netherlands 2·11·149 Posts	Your data save files from that list, which you need to continue the tests, are p332199319 (I think this one is the 100M digit LL test), p01C0989, p3W73763. The same names with .bu and .bu2 extensions are normally slightly older backup versions of the data save files from those tests. Of course, everything now depends on when those data files were made, and whether they've been overwritten by other versions when your tests restarted. So you need to look at the dates and times of last alteration: use the -l extension of the ls command ("ls -l") in Linux to show these times. In worktodo.txt you can see what is currently being worked on (or was being worked on the last time mprime was running, assuming it had access to the file then). You need to make sure that this file contains what you want it to before running mprime. Tip: make copies of what you have and store them elsewhere before you do anything else, certainly before you start mprime again!

2013-03-07, 13:54	#4
obscure "Obscure" Mar 2013 Vienna Austria 1 Posts	Sorry, no older backups found. I should have stopped immediately after seeing something went wrong since the bu2 file is overwritten one hour later. I hoped, some results were stored on the server, not to loose all of the work... What if I unreserve an Exponent, is my work lost then? I was not aware that root cannot read or execute a file but overwrite it as I think this is what really happened. Thank You for help, Obscure

2013-03-08, 11:10	#5
Brian-E "Brian" Jul 2007 The Netherlands CCE₁₆ Posts	This must be so frustrating for you. No, the server simply keeps track of how far through the work you are as reported by your machine, and as for the data processed in your test it only stores the final result. Even then, in the case of LL tests, the server in fact only stores the last 64 bits of the final residue. The data files on your own machine contain the full residue throughout the duration of the test, and this residue is megabytes long. If you unreserve an exponent on the server without "telling" mprime on your machine (worktodo.txt is the master file here), mprime will continue the work. This is not, of course, a good idea: the exponent will at some point be assigned to someone else and you will be doing work which is not yours to do. If you unreserve it more cleanly using the menu system in mprime, or altering the worktodo.txt yourself after manually unreserving on the server, then I think mprime will leave the data files alone (but back them up first to be sure!): but again, if you've unreserved the exponent on the server you should not be continuing the work anyway.

2013-03-08, 11:25	#6
Xyzzy Aug 2002 2168₁₆ Posts	You are not the first to lose work, so please do not feel too bad! We keep a save file for every 1e6 iterations. http://www.mersenneforum.org/showthread.php?p=120849